Major FPS drops with i5-4690k and 980 Ti

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

TheEternal

Reputable
Mar 31, 2014
125
0
4,680
I recently got a 980 Ti to pair with my i5-4690k...neither are overclocked.

I started up Borderlands 2 and I noticed some extremely strange performance. I'm playing on 1080p with max settings on a 144Hz G-sync monitor with a framecap of 135...not very often does my fps hit that cap, but my GPU and CPU never seem fully utilized.

Here's a video I recorded with some gameplay as an FPS test - note that 5:00-5:10 has pretty much the worst drop, with FPS going all way down to 20.

I think I've narrowed the problem down to PhysX, but why would this be the case? I would assume that a 980 Ti would be able to handle maxing PhysX and every other setting in a 3-year-old game with hardly any drops...especially since neither the GPU nor CPU seem to be hitting full load and are staying cool.

Any advice/suggestions would be appreciated!

PC Specs:

Gigabyte G1 980 Ti 6GB
i5-4690k
16GB DDR3
AsRock Pro3 motherboard
OCZ 750W PSU
Windows 10
GeForce Windows 10 Driver (353.62)
 


Well the point of the analogy was that the CPU can't just tick more, the software has to tell it to or the thread has to speed up, and that's based on kernel tick rate and clockrate. Your computer game tells the GPU to draw a factor of these frames as it can to "smooth" motion and make things less jittery. If only your CPU was drawing it would look more like still-photos on a camera. The GPU is predicting where things are by taking the time it took to draw each frame and multiplying it by the 2D/3D Velocity of objects or vectors.

When you're idling at your desktop your GPU could easily draw 1000 times more than it needs to, but there would be no point. Keep in mind your GPU is drawing multiple times per second. So in one second it may need to draw 135 frames and in another it may need to draw 75, but on your screen you're only seeing an average of 105.

There's times when the FPS drops and there's an obvious lag and that's not supposed to happen. There's other times when FPS drops intentionally.
 
Let me make that simpler. Let's say I (Video Game) have a cameraman (CPU), and an artist (GPU). Let's also say I throw a ball and it takes 10 seconds to reach point B from point A. During that 10 seconds the cameraman took 13 shots or "ticks".

Well 13 FPS is really jittery and the ball is blurry, so I ask an artist to draw 70 pictures and arrange them where they should be to create a more fluid video. Now I have 73 FPS and the picture looks perfect.

The GPU is drawing pictures in-between shots rather than after-the-fact. In that sense your GPU doesn't really know how many interpolated frames to draw in-between ticks and has to wait for the CPU to tell it when to stop. Your GPU isn't constantly trying to draw 135 frames and pumping the juice to the max wouldn't necessarily help either.

If you have 10 workers (stream processors) each trying to make their own drawing of the ball, they would have to work together so they don't draw the same picture, and 20 more guys wouldn't really help them, they would be standing around waiting for instruction. If you have 30% of your GPU working on the same picture, well, 10 more percent of that utilization isn't going to help.

You don't want 100% of your GPU being utilized because that means it has nothing free to draw a new scene. You want workers ready to go to do new tasks, not sitting around working on the same task that's already being done.

This is of course an incredible simplification of the process and I don't write GPU drivers.
 

Extremely hard to understand what you're trying to say, but if you're trying to blame this on my GPU I'd say it's not that. Runs fine 99% in other games, like Witcher 3. A lot of the negative reviews seem to be either from damage out of the box or coil whine, which my card has neither of.
 




Thanks, I'll plan to do this, then. Two quick questions, though:

1) Do you recommend I try DDR3-2133 with lower DRAM timings or DDR3-2400 with higher timings first? And which would provide more of a benefit?

2) How should I test the overclock and how will I know if it's unstable or successful? Should I run memtest86+ or just boot and start playing?

I appreciate the help!
 
please try an AMD card, like R9 390 8GB or if you want to go high end (since you have a 980ti), try R9 Fury or Fury X and see if you game is smooth under 15.7.1 drivers

whoever is saying this is "RAM" problem is crazy (you are wasting your time messing with RAM as it has nearly 0% effect on game performance)
 
Thanks Blueberries...makes a bit more sense now. I guess I was thinking the CPU could and would work as much as it needed to on its own to get the job done rather than have to be told when to.
 


I just upgraded from my AMD 7950 to my Nvidia 980 Ti...

How would switching back to AMD, which doesn't have any PhysX support, do me any good? The main problem has pretty much been narrowed down to PhysX... Whether that's a RAM problem or just overall unoptimization I don't know, but I do know it wouldn't be any better with AMD since my card runs fine in pretty much every other game. I was mostly trying to figure out if there was a bottleneck for some reason that I was unaware of.

However, the benchmarks that gskill support posted clearly show an improvement in fps when increasing RAM frequency, so whether RAM has anything to do with my original post or not, I'm still willing to work on my system and get some (free) bonus performance.

 


You mean those really memory-intensive applications that require dedicated GPUs with equally as much VRAM?
 
please show proof of your claims and how they relate to FPS drops (if any)



 
ok, if it's a Physx issue, maybe you can add a 2nd video card like a 960 or 750ti and set it as dedicated Physx only card in the driver?



 


Tell me, how much RAM do you recommend for playing video games? Is it more than 4GB?

Okay then.

Now set your RAM to 800Mhz and play your favorite game and tell me if you notice a difference.

Great, did you do that?

You're officially on the level of everyone else in this thread.
 
What is the best memory for Haswell? From DDR3-1333 to DDR3-2933: performance scaling test:
http://www.xbitlabs.com/articles/memory/display/haswell-ddr3_7.html#sect0





 


So what you're telling me is that in their tests the speed of the memory made no difference and if your memory was operating below what it's rated, misbehaving, failing or had some hardware issue it wouldn't affect you playing video games.

Because I just cannot even comprehend how you came to such a conclusion.
 
he can try playing with Physx off or using a dedicated Physx card like 750/750ti




 
DDR3-2400 should provide more benefit, but you can test both settings to see which performs better. It doesn't hurt to know exactly what's going on and what is actually better.

For stability testing, you can run a quick stress test like Prime95, make sure the system is stable, then you can go right in to gaming to see how it goes.





 


Well, I overclocked it to 1.65V with DDR3-2133 with the timings you suggested (10 - 12 - 12 - 31), and it seemed stable...so I went ahead and OC'd to DDR3-2400 and the timings for that you suggested (11 - 14 - 13 - 31) and everything still seems stable. I ran a few in-game benchmarks: some see no improvement, some see marginal improvement (2-3fps avg, which could just be error), but some games like Bioshock Infinite, Batman: Arkham City, and Thief saw 11-13 more average fps, which is a pretty significant improvement.

I'll test it out a bit more with some gameplay and make sure nothing crashes on me or anything, but would you suggest I continue to increase the frequency or bring down the timings at all?
 
These two dudes saying it's the 980 ti and to try amd lol....That's pretty funny there. The one dude says a 390vs a 980 ti...run along with that. It's Borderlands 2...I don't know how the 980 ti should perform with that game because I don't have it but there are a ton of people, even with titanx that have problems running that game....even in sli. And it absolutely has to do with physx and the unoptimizedness of that game.
 
DRAM frequency is at a good point, it gets a little more tricky overclocking to the next multiplier. Same with timings, unless u plan to increase voltage, it may be near limits as well, but if you feel more comfortable now feel free to do some tweaking to see what the system is capable of. Can even consider some CPU OC..

So you witnessed first hand of the wide range of FPS results depending on game, what do you think of actual game play? There may be a greater range of improvement since you only mentioned avg FPS, improvement in min FPS for example could give the game a completely diff feel. Just wondering if you noticed such change with certain games..
 


CPU OC will probably be my next step once I can find a nice deal on a good cooler...running Prime95 for more than a minute on stock cooler and speeds was taking my CPU to temperatures I'm not comfortable with and that I've never seen in actual gameplay.

As far as gameplay it's hard to say for sure because I haven't had much time with the OC yet, but Borderlands 2 (the game this post originally referred to) unfortunately did not seem much if any better...with PhysX on I was still seeing pretty frequent drops into the 30s.

However, Battlefield 4 doesn't have a benchmark tool so I didn't measure that before and after, but going off of reference from the last time I played, it seemed to run with much smoother framerates. I played for probably about an hour or so, and fps was almost always 120+, never dropped below 100, and often sat at my 135fps cap. Could just be the difference between the maps I was playing, but I definitely remember seeing frame drops to the 80s and 90s fairly frequently before, which I did not experience this time.

Last question: are there any warning signs of unstable RAM other than a bluescreen? I'll take your advice and stop at the DDR3-2400 OC, but I'd just like to know what to look for in case something comes up.

Thanks for your help, I really appreciate it!
 


Yeah, it was a good laugh. I was 99% sure it had nothing to do with my GPU since it runs plenty of other games perfectly fine. If it had anything to do with my system, I'd have guessed just a CPU bottleneck, which does seem to be the case at some points in this game, but the major drops I was referring to is definitely PhysX and overall unoptimization of the game.
 
An overclock might help you some but I don't think it will be your saving grace for that game. It really is borderlands 2, I've seen a ton of threads with people having issues with that game and physx and fps drops and all kinds of stuff, even with high end cards in sli.

You will definitely need a cooler to get any decent overclock, that stock cooler is just barely good enough for your chip at stock speeds let alone any overclock. The 4690K is such a strong chip too once it's overclocked.

Battlefield 4 will vary depending on map too, some I get 150-160FPS at 1440P and others around 120 or so...same with 4K...some 80-90 FPS, others around 60-70.
 


Noticably slower boot times, devices not working normally (mouse not being responsive), applications hanging or loading slowly, computer freezing or crashing, etc.

Usually if there's a problem with your RAM it'll be obvious when using your web browser, it'll take a long time to launch or load web pages.
 
The truth is borderlands 2 has terrible physx performance. its not your PC its the game. You can try load up a different game and you'l see the others run fine but borderlands.I have a i7-4790k and a gtx 770 and with physx on i get 45 fps and dip to 20 allot with phys-x on at max and like most people said the sprite effects or the small particles witch are from phys-x and are enhanced they take a major performance chunk off your gpu.
 


Oh, when I mentioned CPU overclocking, I was mostly talking about as a system improvement for my games overall.. I get MOSTLY good performance in Borderlands 2 if I turn PhysX off, but viewing distance still seems very unoptimized, so there are some drops when looking from areas where you can see a lot of the map. I'll just have to decide if the extra bling PhysX adds to the game is worth the fps drops or not...

And for BF4, I experienced similar situations as you, but my experience today after overclocking my RAM did seem smoother, but that could either be just the luck of the maps and/or placebo effect.